Pith. sign in

REVIEW 2 major objections 4 minor 82 references

Sparse-View Surface Reconstruction using Gaussian Splatting through High-Confidence Depth Propagation with Normal Priors

T0 review · 2 major / 4 minor · reviewed 2026-07-12 · grok-4.5

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Normal-guided depth propagation turns sparse views into high-fidelity 3D surfaces with Gaussian splatting.

desk verdict Solid engineering advance for sparse-view 3DGS surfaces: normal-guided high-confidence depth propagation plus abnormal-edge smoothing beats recent baselines on DTU/TNT, with honest ablations and an admitted texture-less failure mode that does not sink the result. read the letter →

arxiv 2607.03765 v1 pith:N26UGR4D submitted 2026-07-04 cs.CV

Liang Han , Bangcai Wei , Junsheng Zhou , Yu-Shen Liu , Zhizhong Han This is my paper
classification cs.CV
keywords sparse-viewreconstruction3DGaussianSplattingdepthpropagationnormalpriorssurfaceedge-awareregularization
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Reconstructing accurate surfaces from only a few camera views is hard because depth clues are scarce and the discrete Gaussians used by modern renderers leave gaps and jumps. This paper shows that the problem can be solved without large-scale pre-training by first locating the few places where multi-view geometry and color already agree, then using a monocular normal prior to push those reliable depths outward across local planes. A second, edge-aware smoother removes the remaining depth discontinuities that arise where the Gaussians are still under-constrained. On standard sparse-view benchmarks the resulting meshes are smoother and more complete than those of both generalizable networks and other Gaussian-based optimizers, while training still finishes in minutes on a single GPU. The practical payoff is that high-quality geometry becomes available from casually captured photo sets rather than dense, carefully planned scans.

What carries the argument

Normal-guided depth propagation: high-confidence depth pixels (product of geometric reprojection error and photometric NCC) are iteratively extended to neighbors via the local-plane relation d_i = (n_j·r_j)/(n_i·r_i) d_j, expanding the confidence mask and supplying a depth-supervision loss that is later complemented by an abnormal-edge smoother.

What would settle it

On a texture-less planar region that is only partially visible in the sparse views, measure whether the propagated depth still matches a laser-scanned ground-truth surface after the confidence mask has expanded; systematic deviation larger than the method’s reported mean Chamfer would falsify the claim that the anchors remain reliable.

Watch

Extended reading notes

Core claim

High-fidelity surfaces can be recovered from sparse views by treating multi-view consistent depths as anchors and propagating them under normal guidance, then regularizing only the abnormal depth edges that remain; the combination yields lower Chamfer distance and higher F1 than prior scene-specific and generalizable methods on DTU and Tanks-and-Temples.

Load-bearing premise

The multi-view consistency scores used to mark high-confidence anchors stay trustworthy enough that the depths they seed do not introduce large geometric errors when they are propagated.

Editorial extensions

If this is right

  • Casually captured three-view photo sets become sufficient for complete, smooth meshes without dense scanning or multi-day pre-training.
  • Existing monocular normal estimators can be reused as geometric regularizers rather than merely as soft losses, expanding their utility in multi-view pipelines.
  • Gaussian-based optimizers gain an explicit mechanism for filling low-texture and occluded regions that previously produced fragmented surfaces.
  • Training remains under ten minutes and three gigabytes of memory, keeping the approach practical for interactive reconstruction tools.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The same propagation-plus-edge-smoothing pattern could be applied inside other discrete primitive representations (surfels, point clouds) that suffer from under-constrained depth.
  • If a future monocular normal model improves accuracy on texture-less surfaces, the method’s remaining local distortions should shrink without any change to the optimization code.
  • Combining the confidence-propagation idea with a lightweight learned depth prior might further reduce dependence on the initial multi-view consistency cue.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

2 major / 4 minor

Summary. The paper proposes DP-GS, a scene-specific 3D Gaussian Splatting pipeline for high-fidelity surface reconstruction from sparse input views. Its core technical contributions are (i) a normal-guided depth-propagation loss that first identifies high-confidence depth via multi-view geometric and photometric consistency (Eqs. 1–5) and then propagates those depths under monocular normal priors (Eqs. 6–11), and (ii) an abnormal-depth-edge-aware smoothing regularizer that suppresses discontinuities arising from discrete Gaussians. The method is trained with a composite loss (Eq. 14) that also includes RGB and normal-prior terms, and is evaluated primarily by Chamfer distance on 15 DTU scenes with three small-overlap views and by F1 score on TNT scenes with 5/10/20 views, claiming state-of-the-art numbers (mean CD 1.02 / 0.85 with MASt3R initialization) together with competitive runtime (~10 min, 3 GB).

Significance. If the reported gains hold under broader conditions, the work supplies a practical, minutes-scale alternative to both slow NeRF-based sparse-view reconstructors and pre-training-heavy generalizable methods, while remaining fully scene-specific. The explicit ablations (Table 4), iteration and normal-estimator sensitivity studies (Tables 5–6), efficiency table, and candid Limitation paragraph are strengths that make the empirical claims falsifiable and reproducible; a public project page further aids verification. The approach therefore advances the sparse-view surface-reconstruction literature even if subsequent work must harden the confidence cue.

major comments (2)
  1. [§3.1 Eqs. (1)–(5), (8)–(11); Limitation; Table 1; Table 4] The headline mean-CD numbers in Table 1 rest on the binary confidence mask M_C = {C > τ} that gates L_dp (Eqs. 8–11). C itself is the product of multi-view geometric and photometric consistency (Eqs. 1–5). The Limitation section explicitly states that these cues “can become unreliable in texture-less regions, leading to inaccurate confidence estimation” and that “erroneous depths may be propagated.” Because DTU and TNT contain large low-texture planar surfaces, the reported gains may be inflated by lucky high-confidence anchors rather than by a mechanism that is robust precisely where sparse-view geometry is hardest. Table 4 only removes L_dp wholesale; it never isolates the effect of deliberately corrupting or removing the confidence anchors on texture-less patches. A controlled stress test (or at least a quantitative breakdown of CD on high- vs. low-texture regions) is therefore requir
  2. [§3.1 Eq. (6); §3.2; Fig. 5] The depth-propagation formula (Eq. 6) and the subsequent iterative averaging (Eqs. 9–10) rest on a local planar assumption (identical normals). While the abnormal-edge regularizer (Eq. 12) mitigates some resulting discontinuities, the paper never quantifies how often the planar assumption is violated under the three-view small-overlap regime, nor how sensitive final CD is to normal-prior error on curved or occluded surfaces. An additional ablation that injects controlled normal noise or that measures propagation error versus geodesic distance from high-confidence seeds would clarify the practical radius of reliable supervision.
minor comments (4)
  1. [Fig. 6 caption; §1; Table 1 caption] Several figure captions and body sentences contain obvious typos (“sufaces”, “degenerates significantly”, “the # indicates o The # indicates”). A careful proof-reading pass is needed.
  2. [§4.4; Fig. 8] The self-captured experiments (Fig. 8) are purely qualitative; reporting even a simple multi-view photometric consistency score or a relative depth-error metric would make the claim of “best reconstruction performance” more concrete.
  3. [§4.2] Hyper-parameter values (τ = 0.6, t = 10, λ1 = 0.5, au2 = 0.03, λ3 = 0.1) are stated once; a short sensitivity plot or table for the loss weights would help readers reproduce the exact operating point.
  4. [§2.2; Table 4] In the related-work discussion of monocular normal priors, the distinction between “merely applying regularization to the rendered normal” and the proposed propagation is clear, yet a one-sentence quantitative comparison (e.g., CD of a pure L_normal baseline versus full DP-GS) already appears in Table 4 and could be cross-referenced earlier for clarity.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: empirical 3DGS optimization pipeline with external monocular normal prior and multi-view consistency losses, evaluated against independent ground-truth meshes.

full rationale

The paper presents a scene-specific 3D Gaussian Splatting optimization method whose core components (depth-confidence mask C = C_geo · C_pho from multi-view geometric/photometric consistency, normal-guided planar depth propagation di = (nj rj / ni ri) dj, L_dp, abnormal-edge L_ds, and L_normal) are standard regularizers applied during training. The reported Chamfer distances (Table 1) and F1 scores (Table 2) are obtained by comparing the final extracted meshes against external DTU/TNT ground-truth geometry; none of these numbers is algebraically forced by a fitted constant, a self-defined quantity, or a uniqueness claim imported from the authors’ prior work. Self-citations appear only as baselines or related-work references and are not load-bearing for the derivation. The normal prior is taken from an independent off-the-shelf network (Metric3Dv2). The Limitation section candidly notes failure modes of the confidence cue, confirming that the method is not tautological. Consequently the derivation chain is self-contained and non-circular.

Assumptions & free parameters 3 free parameters · 3 assumptions · 1 invented entities

The central performance claim rests on a small set of hand-chosen thresholds and loss weights, the local-planar propagation model, and the assumption that multi-view consistency plus a monocular normal network supply usable geometric supervision under sparse views.

free parameters (3)
  • confidence threshold τ = 0.6
    Hard threshold that decides which rendered depths become propagation anchors (Eq. 8); set to 0.6 without cross-validation reported.
  • propagation iterations t = 10
    Number of neighborhood averaging steps; ablated but finally fixed at 10.
  • loss weights λ1, λ2, λ3 = 0.5, 0.03, 0.1
    Balance terms for depth-propagation, edge-smoothing and normal losses (Eq. 14); chosen by hand.
assumptions (3)
  • domain assumption Local surface patches are planar so that depth can be transferred along the normal (Eq. 6).
    Invoked throughout §3.1; fails at sharp edges and high-curvature regions.
  • domain assumption Multi-view geometric + photometric consistency (NCC) is a reliable indicator of depth accuracy under sparse views.
    Defines the confidence map C (Eqs. 1–5); the Limitation section admits failure in texture-less areas.
  • domain assumption A monocular normal estimator (Metric3Dv2) supplies sufficiently accurate surface orientation for propagation guidance.
    Used as the sole geometric prior; ablation with weaker normals (Omnidata) still works but degrades CD.
invented entities (1)
  • abnormal depth edge mask
    purpose: Isolates depth discontinuities that do not coincide with normal edges so that a coplanarity loss can be applied only there.
    Constructed by multiplying depth-edge and inverted normal-edge masks; no independent geometric meaning outside the proposed regularizer.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Sparse-View Surface Reconstruction using Gaussian Splatting through High-Confidence Depth Propagation with Normal Priors." pith.science (2026). https://pith.science/paper/N26UGR4D

@misc{pith2026260703765,
  author       = {Pith},
  title        = {Pith review of: Sparse-View Surface Reconstruction using Gaussian Splatting through High-Confidence Depth Propagation with Normal Priors},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/N26UGR4D}},
  note         = {Machine review of arXiv:2607.03765}
}
read the original abstract

3D reconstruction from sparse views is a challenging task in 3D computer vision. Recent studies on 3D Gaussian Splatting (3DGS) have achieved remarkable results with sparse views in novel view synthesis, yet reconstructing high-quality geometric surfaces from sparse views remains a challenge, due to the limited geometry clues and the discreteness of Gaussians. In this paper, we propose a novel 3DGS-based method for high-fidelity surface reconstruction from sparse views. Our key insight is to introduce a normal-guided depth propagation approach, which can extend depth information from high-confidence regions to constrain the depth in low-confidence areas. Additionally, we propose an abnormal depth edge-aware regularization to address depth discontinuities caused by the discreteness of Gaussians. Extensive experiments on DTU and Tanks-and-Temples datasets demonstrate that our method outperforms the state-of-the-art methods in sparse view surface reconstruction. Project page: https://hanl2010.github.io/DP-GS.

Figures

Figures reproduced from arXiv: 2607.03765 by the authors.

Figure 1
Figure 1. Reconstructed surfaces from three small-overlap views of the DTU dataset. Compared with state-of-the-art 3DGS-based methods FatesGS [19] and MAtCha [13], Our method produces more faithful surface reconstruction. Two types of solutions have been proposed for addressing this issue: gen￾eralizable methods [33, 34, 41, 43, 62] and scene-specific optimization methods [15, 19, 20, 56, 69, 71]. Generalizable methods requir… view at source ↗
Figure 2
Figure 2. The overview of our method. (a) We leverage a normal prior to propagate depth information from high-confidence regions to low-confidence regions, effectively constraining the depth in regions with lower confidence. (b) The edge-aware regular￾ization for abnormal depths is employed to eliminate depth discontinuities caused by the insufficient constraints of the unstructured Gaussian. In addition, we also use color lo… view at source ↗
Figure 3
Figure 3. Illustration of depth confidence cal￾culation. Depth Confidence Map As shown in [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: Illustration of the depth propaga￾tion. Depth Propagation Our goal of depth propagation is to leverage nor￾mal priors to obtain a broader range of accurate depth predictions on planes, which contributes to eliminat￾ing the uncertainty of Gaussian depth at low-confidenc…
Figure 5
Figure 5. Figure 5 [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: Visual comparison of reconstruction with 3 small-overlapping views on the DTU dataset. 4 Experiments 4.1 Datasets We evaluate the performance of our method on three datasets: DTU [22], Tanks and Temples (TNT) [25] and 5 scenes captured by ourselves. We follow previous …
Figure 7
Figure 7. Figure 7: Visual comparison of reconstruction with sparse views on the TNT dataset. methods, such as MonoSDF [71], NeusSurf [20], SparseCraft [69], SparseRecon [15], MAtCha [13] and FatesGS [19]. We also compare our approach with the latest dense-view reconstruction methods base…
Figure 8
Figure 8. Figure 8: Visual comparison on self-captured dataset with 3 input views [PITH_FULL_IMAGE:figures/full_fig_p013_8.png]
Figure 9
Figure 9. Figure 9: Visualization of ablation results on Scan 24 in the DTU dataset. O.D. stands for the opacity decay strategy. To verify the contribution of ei￾ther the depth propagation constraint loss Ldp or the depth edge smooth￾ing loss Lds, we add each of them to the normal loss an…
Figure 10
Figure 10. Figure 10: provides a clearer illustration of the depth anomalies, which lead to mesh fragmentation. The comparison shows that Lds significantly improves the quality of the reconstructed geometry. (a) w/o (b) w/ Rendered Depth Rendered DepthMesh Mesh [PITH_FULL_IMAGE:figures/fu…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

82 extracted references · 2 linked inside Pith

  1. [1]

    In: British Machine Vision Con- ference (BMVC) (2022) 5

    Bae, G., Budvytis, I., Cipolla, R.: Irondepth: Iterative refinement of single-view depth using surface normal and its uncertainty. In: British Machine Vision Con- ference (BMVC) (2022) 5

    2022

  2. [2]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Barron, J.T., Mildenhall, B., Verbin, D., Srinivasan, P.P., Hedman, P.: Mip- NeRF 360: Unbounded anti-aliased neural radiance fields. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 5470–5479 (2022) 4

    2022

  3. [3]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Cai, B., Huang, J., Jia, R., Lv, C., Fu, H.: NeuDA: Neural deformable anchor for high-fidelity implicit surface reconstruction. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 8476–8485 (2023) 4

    2023

  4. [4]

    IEEE Transactions on Visualization and Computer Graph- ics31, 1–12 (01 2024) 1, 4, 6, 9, 10, 11

    Chen, D., Li, H., Ye, W., Wang, Y., Xie, W., Zhai, S., Wang, N., Liu, H., Bao, H., Zhang, G.: PGSR: Planar-based gaussian splatting for efficient and high-fidelity surface reconstruction. IEEE Transactions on Visualization and Computer Graph- ics31, 1–12 (01 2024) 1, 4, 6, 9, 10, 11

    2024

  5. [5]

    In: European Conference on Computer Vision

    Chen, Y., Xu, H., Zheng, C., Zhuang, B., Pollefeys, M., Geiger, A., Cham, T.J., Cai, J.: MVSplat: Efficient 3D gaussian splatting from sparse multi-view images. In: European Conference on Computer Vision. pp. 370–386. Springer (2024) 2, 4

    2024

  6. [6]

    In: ACM SIGGRAPH 2024 Conference Proceed- ings (2024) 1, 4

    Dai, P., Xu, J., Xie, W., Liu, X., Wang, H., Xu, W.: High-quality surface recon- struction using gaussian surfels. In: ACM SIGGRAPH 2024 Conference Proceed- ings (2024) 1, 4

    2024

  7. [7]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Darmon, F., Bascle, B., Devaux, J.C., Monasse, P., Aubry, M.: Improving neural implicit surfaces geometry with patch warping. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 6260–6269 (2022) 1, 4

    2022

  8. [8]

    In: Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition

    Deng, K., Liu, A., Zhu, J.Y., Ramanan, D.: Depth-supervised NeRF: Fewer views and faster training for free. In: Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition. pp. 12882–12891 (2022) 4

    2022

Show all 82 references
  1. [9]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Dogaru,A.,Ardelean,A.T.,Ignatyev,S.,Zakharov,E.,Burnaev,E.:Sphere-guided training of neural implicit surfaces. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 20844–20853 (2023) 4

    2023

  2. [10]

    In: 2025 International Conference on 3D Vision (3DV)

    Duisterhof, B.P., Zust, L., Weinzaepfel, P., Leroy, V., Cabon, Y., Revaud, J.: Mast3R-SfM: a fully-integrated solution for unconstrained structure-from-motion. In: 2025 International Conference on 3D Vision (3DV). pp. 1–10. IEEE (2025) 9

    2025

  3. [11]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision

    Eftekhar, A., Sax, A., Malik, J., Zamir, A.: Omnidata: A scalable pipeline for making multi-task mid-level vision datasets from 3D scans. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 10786–10796 (2021) 15

    2021

  4. [12]

    Advances in Neural Information Processing Systems35, 3403–3416 (2022) 1, 4

    Fu, Q., Xu, Q., Ong, Y.S., Tao, W.: Geo-NeuS: Geometry-consistent neural implicit surfaces learning for multi-view reconstruction. Advances in Neural Information Processing Systems35, 3403–3416 (2022) 1, 4

    2022

  5. [13]

    In: Proceedings of the Computer Vision and Pattern Recognition Conference

    Guédon, A., Ichikawa, T., Yamashita, K., Nishino, K.: MAtCha Gaussians: At- las of Charts for High-Quality Geometry and Photorealism From Sparse Views. In: Proceedings of the Computer Vision and Pattern Recognition Conference. pp. 6001–6011 (2025) 2, 5, 9, 10, 11, 12, 15

    2025

  6. [14]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Guédon, A., Lepetit, V.: SuGaR: Surface-aligned gaussian splatting for efficient 3D mesh reconstruction and high-quality mesh rendering. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 5354– 5363 (2024) 1, 4 Sparse-View Surface Recons...

    2024

  7. [15]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision

    Han, L., Zhang, X., Song, H., Shi, K., Liu, Y.S., Han, Z.: SparseRecon: Neural implicit surface reconstruction from sparse views with feature and depth consis- tencies. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 28514–28524 (2025) 2, 4, 9,...

    2025

  8. [16]

    In: Advances in Neural Information Processing Systems (2024) 2, 4, 10, 13

    Han, L., Zhou, J., Liu, Y.S., Han, Z.: Binocular-guided 3D gaussian splatting with view consistency for sparse view synthesis. In: Advances in Neural Information Processing Systems (2024) 2, 4, 10, 13

    2024

  9. [17]

    IEEE Transactions on Pattern Analysis and Machine Intelligence (2024) 4, 10, 15

    Hu, M., Yin, W., Zhang, C., Cai, Z., Long, X., Chen, H., Wang, K., Yu, G., Shen, C., Shen, S.: Metric3Dv2: A versatile monocular geometric foundation model for zero-shot metric depth and surface normal estimation. IEEE Transactions on Pattern Analysis and Machine Intelligence ...

    2024

  10. [18]

    In: ACM SIGGRAPH 2024 Conference Proceedings (2024) 1, 4, 9, 11

    Huang, B., Yu, Z., Chen, A., Geiger, A., Gao, S.: 2D gaussian splatting for geomet- rically accurate radiance fields. In: ACM SIGGRAPH 2024 Conference Proceedings (2024) 1, 4, 9, 11

    2024

  11. [19]

    In: Proceedings of the AAAI Conference on Artificial Intelli- gence (2025) 2, 4, 5, 9, 10, 11, 12

    Huang, H., Wu, Y., Deng, C., Gao, G., Gu, M., Liu, Y.S.: FatesGS: Fast and accurate sparse-view surface reconstruction using gaussian splatting with depth- feature consistency. In: Proceedings of the AAAI Conference on Artificial Intelli- gence (2025) 2, 4, 5, 9, 10, 11, 12

    2025

  12. [20]

    In: Proceedings of the AAAI Conference on Artificial Intelligence

    Huang, H., Wu, Y., Zhou, J., Gao, G., Gu, M., Liu, Y.S.: NeuSurf: On-surface priors for neural surface reconstruction from sparse input views. In: Proceedings of the AAAI Conference on Artificial Intelligence. vol. 38, pp. 2312–2320 (2024) 2, 4, 9, 10, 11

    2024

  13. [21]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision

    Jain, A., Tancik, M., Abbeel, P.: Putting NeRF on a Diet: Semantically consis- tent few-shot view synthesis implementation. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 5885–5894 (2021) 4

    2021

  14. [22]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Jensen, R., Dahl, A., Vogiatzis, G., Tola, E., Aanæs, H.: Large scale multi-view stereopsis evaluation. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 406–413 (2014) 10, 13

    2014

  15. [23]

    In: ProceedingsoftheIEEE/CVFConferenceonComputerVisionandPatternRecog- nition

    Ke, B., Obukhov, A., Huang, S., Metzger, N., Daudt, R.C., Schindler, K.: Re- purposing diffusion-based image generators for monocular depth estimation. In: ProceedingsoftheIEEE/CVFConferenceonComputerVisionandPatternRecog- nition. pp. 9492–9502 (2024) 4

    2024

  16. [24]

    ACM Transactions on Graphics42(4), 1–14 (2023) 1, 2, 4, 5, 10, 11

    Kerbl, B., Kopanas, G., Leimkühler, T., Drettakis, G.: 3D gaussian splatting for real-time radiance field rendering. ACM Transactions on Graphics42(4), 1–14 (2023) 1, 2, 4, 5, 10, 11

    2023

  17. [25]

    ACM Transactions on Graphics (ToG)36(4), 1–13 (2017) 10, 11

    Knapitsch, A., Park, J., Zhou, Q.Y., Koltun, V.: Tanks and temples: Benchmarking large-scale scene reconstruction. ACM Transactions on Graphics (ToG)36(4), 1–13 (2017) 10, 11

    2017

  18. [26]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Li, J., Zhang, J., Bai, X., Zheng, J., Ning, X., Zhou, J., Gu, L.: DNGaussian: Optimizing sparse-view 3D gaussian radiance fields with global-local depth nor- malization. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 20775–20785 (2024) 2, 4

    2024

  19. [27]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision

    Li, Q., Feng, H., Gong, X., Liu, Y.S.: Learning normals of noisy points by local gradient-aware surface filtering. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 28828–28838 (2025) 4

    2025

  20. [28]

    Advances in Neural Information Processing Systems38, 173867–173886 (2026) 4

    Li, Q., Feng, H., Gong, X., Liu, Y.S.: VA-GS: Enhancing the geometric represen- tation of gaussian splatting via view alignment. Advances in Neural Information Processing Systems38, 173867–173886 (2026) 4

    2026

  21. [29]

    IEEE Transactions on Visualization and Computer Graphics (2025) 4 18 L

    Li, Q., Feng, H., Shi, K., Gao, Y., Fang, Y., Liu, Y.S., Han, Z.: PFF-Net: Patch fea- ture fitting for point cloud normal estimation. IEEE Transactions on Visualization and Computer Graphics (2025) 4 18 L. Han et al

    2025

  22. [30]

    Li, S., Liu, Y.S., Han, Z.: GaussianUDF: Inferring unsigned distance functions through3Dgaussiansplatting.In:ProceedingsoftheComputerVisionandPattern Recognition Conference. pp. 27113–27123 (2025) 4

    2025

  23. [31]

    Computational Visual Media (2025) 4

    Li, X., Tan, X., Zhang, Z., Xie, Y., Ma, L.: Point mask transformer for outdoor point cloud semantic segmentation. Computational Visual Media (2025) 4

    2025

  24. [32]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Li, Z., Müller, T., Evans, A., Taylor, R.H., Unberath, M., Liu, M.Y., Lin, C.H.: Neuralangelo: High-fidelity neural surface reconstruction. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 8456– 8465 (2023) 1, 4

    2023

  25. [33]

    Advances in Neural Information Processing Systems36(2024) 2, 4, 9, 10

    Liang, Y., He, H., Chen, Y.: ReTR: Modeling rendering via transformer for gener- alizable neural surface reconstruction. Advances in Neural Information Processing Systems36(2024) 2, 4, 9, 10

    2024

  26. [34]

    In: European Conference on Computer Vision

    Long, X., Lin, C., Wang, P., Komura, T., Wang, W.: SparseNeuS: Fast generaliz- able neural surface reconstruction from sparse views. In: European Conference on Computer Vision. pp. 210–227. Springer (2022) 2, 4, 9, 10

    2022

  27. [35]

    ACM Trans- actions on Graphics (TOG)43(6), 1–12 (2024) 1, 4

    Lyu, X., Sun, Y.T., Huang, Y.H., Wu, X., Yang, Z., Chen, Y., Pang, J., Qi, X.: 3DGSR: Implicit surface reconstruction with 3D gaussian splatting. ACM Trans- actions on Graphics (TOG)43(6), 1–12 (2024) 1, 4

    2024

  28. [36]

    Computational Visual Media (2025) 4

    Mao, X., Yuan, H., Lu, X., Hamzaoui, R., Gao, W.: PCAC-GAN: a sparse-tensor- based generative adversarial network for 3D point cloud attribute compression. Computational Visual Media (2025) 4

    2025

  29. [37]

    Com- munications of the ACM65(1), 99–106 (2021) 1, 2, 4

    Mildenhall, B., Srinivasan, P.P., Tancik, M., Barron, J.T., Ramamoorthi, R., Ng, R.: NeRF: Representing scenes as neural radiance fields for view synthesis. Com- munications of the ACM65(1), 99–106 (2021) 1, 2, 4

    2021

  30. [38]

    ACM transactions on graphics (TOG)41(4), 1–15 (2022) 4, 5

    Müller,T.,Evans,A.,Schied,C.,Keller,A.:Instantneuralgraphicsprimitiveswith a multiresolution hash encoding. ACM transactions on graphics (TOG)41(4), 1–15 (2022) 4, 5

    2022

  31. [39]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Na,Y.,Kim,W.J.,Han,K.B.,Ha,S.,Yoon,S.E.:UFORecon:Generalizablesparse- view surface reconstruction from arbitrary and unfavorable sets. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 5094–5104 (2024) 4, 9, 10, 11, 12

    2024

  32. [40]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Niemeyer, M., Barron, J.T., Mildenhall, B., Sajjadi, M.S., Geiger, A., Radwan, N.: RegNeRF: Regularizing neural radiance fields for view synthesis from sparse inputs. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 5480–5490 (2022) 4

    2022

  33. [41]

    In: Advances in Neural Information Processing Systems

    Peng, R., Gu, X., Tang, L., Shen, S., Yu, F., Wang, R.: GenS: Generalizable neural surface reconstruction from multi-view images. In: Advances in Neural Information Processing Systems. vol. 36, pp. 56932–56945 (2023) 2, 4, 9, 10

    2023

  34. [42]

    IEEE Transactions on Pattern Analysis and Machine Intelli- gence44(2), 969–984 (2020) 5

    Qi, X., Liu, Z., Liao, R., Torr, P.H., Urtasun, R., Jia, J.: GeoNet++: Iterative geometric neural network with edge-aware refinement for joint depth and surface normal estimation. IEEE Transactions on Pattern Analysis and Machine Intelli- gence44(2), 969–984 (2020) 5

    2020

  35. [43]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Ren, Y., Zhang, T., Pollefeys, M., Süsstrunk, S., Wang, F.: VolRecon: Volume ren- dering of signed ray distance functions for generalizable multi-view reconstruction. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 16685–16695 (2023) ...

    2023

  36. [44]

    In: European Conference on Computer Vision (2016) 11 Sparse-View Surface Reconstruction 19

    Schönberger, J.L., Zheng, E., Pollefeys, M., Frahm, J.M.: Pixelwise view selection for unstructured multi-view stereo. In: European Conference on Computer Vision (2016) 11 Sparse-View Surface Reconstruction 19

    2016

  37. [45]

    arXiv preprint arXiv:2412.15400 (2024) 3, 4, 5

    Shen, Z., Liu, Y., Chen, Z., Li, Z., Wang, J., Liang, Y., Yu, Z., Zhang, J., Xu, Y., Schaefer, S., et al.: SolidGS: Consolidating gaussian surfel splatting for sparse-view surface reconstruction. arXiv preprint arXiv:2412.15400 (2024) 3, 4, 5

    2024 arXiv

  38. [46]

    In: ACM SIGGRAPH 2022 Conference Pro- ceedings

    Sun, J., Chen, X., Wang, Q., Li, Z., Averbuch-Elor, H., Zhou, X., Snavely, N.: Neural 3D reconstruction in the wild. In: ACM SIGGRAPH 2022 Conference Pro- ceedings. pp. 1–9 (2022) 4

    2022

  39. [47]

    In: Proceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition

    Truong, P., Rakotosaona, M.J., Manhardt, F., Tombari, F.: SPARF: Neural radi- ance fields from sparse and noisy poses. In: Proceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition. pp. 4190–4200 (2023) 4

    2023

  40. [48]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision

    Wang, G., Chen, Z., Loy, C.C., Liu, Z.: SparseNeRF: Distilling depth ranking for few-shot novel view synthesis. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 9065–9076 (2023) 4

    2023

  41. [49]

    Wang, J., Liu, Y., Wang, P., Lin, C., Hou, J., Li, X., Komura, T., Wang, W.: GausSurf:Geometry-guided3Dgaussiansplattingforsurfacereconstruction.arXiv preprint arXiv:2411.19454 (2024) 3, 5

    2024 arXiv

  42. [50]

    In: European conference on computer vision

    Wang, J., Wang, P., Long, X., Theobalt, C., Komura, T., Liu, L., Wang, W.: NeuRIS: Neural reconstruction of indoor scenes using normal priors. In: European conference on computer vision. pp. 139–155. Springer (2022) 5

    2022

  43. [51]

    Advances in Neural Information Processing Systems (2021) 1, 4

    Wang, P., Liu, L., Liu, Y., Theobalt, C., Komura, T., Wang, W.: NeuS: Learn- ing neural implicit surfaces by volume rendering for multi-view reconstruction. Advances in Neural Information Processing Systems (2021) 1, 4

    2021

  44. [52]

    Compu- tational Visual Media10(4), 659–669 (2024) 4

    Wang, W., Liu, X., Zhou, H., Wei, L., Deng, Z., Murshed, M., Lu, X.: Noise4Denoise: Leveraging noise for unsupervised point cloud denoising. Compu- tational Visual Media10(4), 659–669 (2024) 4

    2024

  45. [53]

    In: Proceed- ings of the IEEE/CVF International Conference on Computer Vision

    Wang, Y., Han, Q., Habermann, M., Daniilidis, K., Theobalt, C., Liu, L.: NeuS2: Fast learning of neural implicit surfaces for multi-view reconstruction. In: Proceed- ings of the IEEE/CVF International Conference on Computer Vision. pp. 3295– 3306 (2023) 1, 4

    2023

  46. [54]

    Advances in Neural Information Processing Systems 35, 1966–1978 (2022) 1, 4

    Wang, Y., Skorokhodov, I., Wonka, P.: HF-NeuS: Improved surface reconstruction using high-frequency details. Advances in Neural Information Processing Systems 35, 1966–1978 (2022) 1, 4

    1966

  47. [55]

    IEEE Transactions on Pattern Analysis and Machine Intelligence45(1), 852–867 (2022) 4

    Wen, X., Xiang, P., Han, Z., Cao, Y.P., Wan, P., Zheng, W., Liu, Y.S.: PMP- Net++: Point cloud completion by transformer-enhanced multi-step point moving paths. IEEE Transactions on Pattern Analysis and Machine Intelligence45(1), 852–867 (2022) 4

    2022

  48. [56]

    International Conference on Computer Vision (2023) 2, 4

    Wu, H., Graikos, A., Samaras, D.: S-VolSDF: Sparse multi-view stereo regular- ization of neural implicit surfaces. International Conference on Computer Vision (2023) 2, 4

    2023

  49. [57]

    In: Pro- ceedings of the Computer Vision and Pattern Recognition Conference

    Wu, J., Li, R., Zhu, Y., Guo, R., Sun, J., Zhang, Y.: Sparse2DGS: Geometry- prioritized gaussian splatting for surface reconstruction from sparse views. In: Pro- ceedings of the Computer Vision and Pattern Recognition Conference. pp. 11307– 11316 (2025) 2, 4, 9

    2025

  50. [58]

    Wu, T., Wang, J., Pan, X., Xu, X., Theobalt, C., Liu, Z., Lin, D.: Voxurf: Voxel- basedefficientandaccurateneuralsurfacereconstruction.InternationalConference on Learning Representations (2022) 1, 4

    2022

  51. [59]

    Computational Visual Media10(4), 613–642 (2024) 1

    Wu, T., Yuan, Y.J., Zhang, L.X., Yang, J., Cao, Y.P., Yan, L.Q., Gao, L.: Recent advances in 3D gaussian splatting. Computational Visual Media10(4), 613–642 (2024) 1

    2024

  52. [60]

    Han et al

    Xiang, H., Li, X., Cheng, K., Lai, X., Zhang, W., Liao, Z., Zeng, L., Liu, X.: Gaus- sianRoom: Improving 3D gaussian splatting with SDF guidance and monocular 20 L. Han et al. cues for indoor scene reconstruction. In: 2025 IEEE International Conference on Robotics and Automati...

    2025

  53. [61]

    IEEE Transactions on Pattern Analysis and Machine Intelligence 45(5), 6320–6338 (2022) 4

    Xiang, P., Wen, X., Liu, Y.S., Cao, Y.P., Wan, P., Zheng, W., Han, Z.: Snowflake point deconvolution for point cloud completion and generation with skip-transformer. IEEE Transactions on Pattern Analysis and Machine Intelligence 45(5), 6320–6338 (2022) 4

    2022

  54. [62]

    In: Proceedings of the IEEE/CVF International Conference on Com- puter Vision

    Xu, L., Guan, T., Wang, Y., Liu, W., Zeng, Z., Wang, J., Yang, W.: C2F2NeuS: Cascade cost frustum fusion for high fidelity and generalizable neural surface re- construction. In: Proceedings of the IEEE/CVF International Conference on Com- puter Vision. pp. 18291–18301 (2023) 2, 4

    2023

  55. [63]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Yang, J., Pavone, M., Wang, Y.: FreeNeRF: Improving few-shot neural rendering with free frequency regularization. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 8254–8263 (2023) 4

    2023

  56. [64]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Yang, L., Kang, B., Huang, Z., Xu, X., Feng, J., Zhao, H.: Depth Anything: Un- leashing the power of large-scale unlabeled data. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 10371–10381 (2024) 4

    2024

  57. [65]

    Computational Visual Media11(1), 83–101 (2025) 4

    Yang, Y.Q., Guo, Y.X., Xiong, J.Y., Liu, Y., Pan, H., Wang, P.S., Tong, X., Guo, B.:Swin3D:Apretrainedtransformerbackbonefor3Dindoorsceneunderstanding. Computational Visual Media11(1), 83–101 (2025) 4

    2025

  58. [66]

    Advancesin Neural Information ProcessingSystems34,4805–4815(2021) 1, 4

    Yariv, L., Gu, J., Kasten, Y., Lipman, Y.: Volume rendering of neural implicit surfaces. Advancesin Neural Information ProcessingSystems34,4805–4815(2021) 1, 4

    2021

  59. [67]

    ACM Trans- actions on Graphics (TOG)43(6), 1–18 (2024) 15

    Ye, C., Qiu, L., Gu, X., Zuo, Q., Wu, Y., Dong, Z., Bo, L., Xiu, Y., Han, X.: Sta- bleNormal: Reducing diffusion variance for stable and sharp normal. ACM Trans- actions on Graphics (TOG)43(6), 1–18 (2024) 15

    2024

  60. [68]

    Circuits, systems and sig- nal processing28, 819–843 (2009) 6

    Yoo, J.C., Han, T.H.: Fast normalized cross-correlation. Circuits, systems and sig- nal processing28, 819–843 (2009) 6

    2009

  61. [69]

    In: European Conference on Computer Vision

    Younes, M., Ouasfi, A., Boukhayma, A.: SparseCraft: Few-shot neural reconstruc- tion through stereopsis guided geometric linearization. In: European Conference on Computer Vision. pp. 37–56. Springer (2024) 2, 4, 5, 9, 11

    2024

  62. [70]

    Advances in Neural Information Processing Systems37, 129507–129530 (2024) 4

    Yu, M., Lu, T., Xu, L., Jiang, L., Xiangli, Y., Dai, B.: GSDF: 3DGS meets SDF for improved neural rendering and reconstruction. Advances in Neural Information Processing Systems37, 129507–129530 (2024) 4

    2024

  63. [71]

    Advances in Neural Information Processing Systems35, 25018–25032 (2022) 1, 2, 4, 5, 9, 10, 11

    Yu, Z., Peng, S., Niemeyer, M., Sattler, T., Geiger, A.: MonoSDF: Exploring monocular geometric cues for neural implicit surface reconstruction. Advances in Neural Information Processing Systems35, 25018–25032 (2022) 1, 2, 4, 5, 9, 10, 11

    2022

  64. [72]

    ACM Transactions on Graphics43(6), 1–13 (2024) 1

    Yu, Z., Sattler, T., Geiger, A.: Gaussian Opacity Fields: Efficient adaptive surface reconstruction in unbounded scenes. ACM Transactions on Graphics43(6), 1–13 (2024) 1

    2024

  65. [73]

    IEEE Transactions on Pattern Analysis and Machine Intelligence45(7), 8713–8728 (2022) 4

    Yuan, Y.J., Lai, Y.K., Huang, Y.H., Kobbelt, L., Gao, L.: Neural radiance fields from sparse RGB-D images for high-quality view synthesis. IEEE Transactions on Pattern Analysis and Machine Intelligence45(7), 8713–8728 (2022) 4

    2022

  66. [74]

    ACM Transactions on Graphics45(2), 1–14 (2026) 4

    Zhang, B., Fang, C., Shrestha, R., Liang, Y., Long, X.X., Tan, P.: RaDe-GS: Ras- terizing depth in gaussian splatting. ACM Transactions on Graphics45(2), 1–14 (2026) 4

    2026

  67. [75]

    In: European Conference on Computer Vision

    Zhang,J.,Li,J.,Yu,X.,Huang,L.,Gu,L.,Zheng,J.,Bai,X.:CoR-GS:sparse-view 3D gaussian splatting via co-regularization. In: European Conference on Computer Vision. pp. 335–352. Springer (2024) 2, 4 Sparse-View Surface Reconstruction 21

    2024

  68. [76]

    In: Pro- ceedings of the Computer Vision and Pattern Recognition Conference

    Zhang, W., Jia, E.Y.t., Zhou, J., Ma, B., Shi, K., Liu, Y.S., Han, Z.: NeRFPrior: Learning neural radiance field as a prior for indoor scene reconstruction. In: Pro- ceedings of the Computer Vision and Pattern Recognition Conference. pp. 11317– 11327 (2025) 4

    2025

  69. [77]

    Advances in Neural Information Processing Systems37, 101856–101879 (2025) 1, 4

    Zhang, W., Liu, Y.S., Han, Z.: Neural signed distance function inference through splatting 3D gaussians pulled on zero-level set. Advances in Neural Information Processing Systems37, 101856–101879 (2025) 1, 4

    2025

  70. [78]

    IEEE Transactions on Pattern Analysis and Machine Intelligence (2026) 4

    Zhang, W., Wang, C., Shi, K., Liu, Y.S., Han, Z.: VRP-UDF: Towards unbiased learning of unsigned distance functions from multi-view images with volume ren- dering priors. IEEE Transactions on Pattern Analysis and Machine Intelligence (2026) 4

    2026

  71. [79]

    IEEE Transactions on Image Processing32, 2703–2718 (2023) 4

    Zhang, W., Xing, R., Zeng, Y., Liu, Y.S., Shi, K., Han, Z.: Fast learning radiance fields by shooting much fewer rays. IEEE Transactions on Image Processing32, 2703–2718 (2023) 4

    2023

  72. [80]

    In: Proceedings of the Computer Vision and Pattern Recognition Conference

    Zhang, W., Yang, Y., Huang, H., Han, L., Shi, K., Liu, Y.S., Han, Z.: MonoIn- stance: Enhancing monocular priors via multi-view instance alignment for neural rendering and reconstruction. In: Proceedings of the Computer Vision and Pattern Recognition Conference. pp. 21642–2165...

    2025

  73. [81]

    In: European Conference on Computer Vision

    Zhu, Z., Fan, Z., Jiang, Y., Wang, Z.: FSGS: Real-time few-shot view synthesis using gaussian splatting. In: European Conference on Computer Vision. pp. 145–

  74. [82]

    Springer (2025) 2, 4

    2025

Pith tools

Reviewed July 12, 2026 · model on record in the stance chip above.